Gary Marcus

Rank 13 of 47
|
Score 83
4/29/2025, 1:34:17 PM
X
In reply to:
Tyler John
@tyler_m_john
·
259d
@GaryMarcus I think this from @davidad would work but it would cost a zillion dollars

https://www.lesswrong.com/posts/…
Tyler John
@tyler_m_john
·
259d
@GaryMarcus I do think constitutional AI and RLAIF is a good innovation that has a chance of working
Tyler John
@tyler_m_john
·
259d
@GaryMarcus I think you and I are in full... alignment on that
Gary Marcus
@GaryMarcus
·
259d
@tyler_m_john Yes, 2 was awful, but 4 isn’t really aligned, and pure scaling won’t get us there, and I am wondering if there are others kinds of results eg with fine tuning, RLHF, or other techniques that look like they will ever get past a kind of 80:20 of working when something is close
Tyler John
@tyler_m_john
·
259d
@GaryMarcus let me think about it today, and retweet you so my audience of people on the bleeding edge of alignment can comment

Offhand it really is just that GPT-2 was so goddamn awful on any possible way of measuring it

It's possible to prompt it here

https://huggingface.co/openai-co…
Gary Marcus
@GaryMarcus
·
259d
@tyler_m_john I am; so much teaching to the test. But “it feels better” is worse. I am asking about the systematic research that most impresses you.

The statement is part of a public discourse on AI alignment and the effectiveness of different AI models. It engages with a substantive issue in the field of artificial intelligence, specifically discussing the potential and limitations of large language models (LLMs) and the importance of world models and verification.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and is a neutral contribution to the discussion, focusing on technical aspects rather than personal attacks. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by agreeing with the importance of world models and verification, contributing to a constructive dialogue. [+1]
  3. Principle 4:
    I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
    The statement engages in constructive criticism by expressing agreement with a viewpoint and sharing a perspective on the limitations of pure LLMs without resorting to personal attacks. [+1]